 If you take a seat, we can resume with the series of four presentations. There will be shorter presentation, but all questions are welcome. Take a seat. Can you help me? Because if I press this button and this, you say try to do this, but it works. Forward, backward, and the pointer. I think you do not touch anything else. Thank you. OK, so the first one, it will be Eileen Cabrera. She will speak about scenario analysis to reduce the risk of vegetation cover loss due to wildfire in Colombia based on the machine learning approach. Ten minutes, I'll give you a warning. Oh, thank you. Hello, everyone. I want to present you my final project of my bachelor. So it is wildfire scenarios for assessing risk of cover loss in a mega diverse zone. This has been the zone that we have selected. We selected this zone because it has different important characteristics. So how you can see, we have different ecosystems here. So we have about eight ecosystems here. And at the same time, we have a high endemic level. So that's one of the reasons why we have selected this zone. Another essential reason is the high quantity of hotspot we have seen in this part. So we can see that most of the hotspot are located in the south part and in the north part. How you can see. What we want to do with this project is we want to evaluate mitigation strategies. And to get that, what we want to do is to create sensitivity experiments. And so what we did was we had to measure vulnerability and we had to measure the damage. In the case of vulnerability, we evaluate ecological and socioeconomic factors. And in the case of the damage, we evaluate meteorology. And I'll explain you more about this in the next step. In the case of the damage, we use a machine learning model. So we use artificial neural networks. This is the structure of the network, how you can see. And you can see that we are based on the convolutional neural network. Those that you can see are the inputs. And we have evaporation, wind, wind speed, precipitation, solar radiation, a leaf area index for the high vegetation, but also for the low vegetation, the land cover and the skin or the server content. But another important thing that we added was different FWI indexes. And the output was the air temperature, which is some matrix. This has been the results that we had obtained. So we can see the observations and the prediction. We can see that it works in different parts. So it follows the results, follows the trend. But sometimes it has an overestimation. But we can see that the trend is good. It captures the spatial distribution. We can see the results in another way. So here in that part, we have the mean, the mean of all the some because we had different pixels. And we can see that it follows the trend. And this is the example of just one pixel. And we can see that it also follows that trend. And at the same time, we can see the evaluation of this model. So we have the bias, the RMSE, and R. And it shows good results. But we are improving this model. And for example, in that specific part, you can see that those are the points where we have wildfires. So even though it doesn't have the exact, the specific values that we need, it shows good results. And in the case of vulnerability, we base our method on Paramore-Rochet's methodology. And we evaluate different things in this part. We have ecological vulnerability, but we also have socioeconomic vulnerability. In the case of ecological vulnerability, we evaluated net susceptibility, in which we took into account different characteristics of the vegetation and the ecosystems. So we took into consideration the full duration, full load, for influence on ecosystems, temperature, and precipitation. And in the other part, we have ecosystems at risk, because maybe ecosystems are another kind of problems. For example, maybe the land use is not the right land use. For the rare land use, it's not the correct. In the other part, we have socioeconomic vulnerability. Here, we evaluate we, which are the areas that are close to the urban centers. So maybe those areas, we have vegetation, are present more risk because of the anthropogenic activities. And in the other side, we have response capacity, because municipalities have different activities that they can implement to face these wildfires. So maybe they don't have enough resources or things like that, like those. And what we are about to do is Monte Carlo simulations, because we want to add uncertainty to the model. So what we want to do is to modify the entry of the machine learning model, adding Monte Carlo simulations. And the other part, which is the final part that we want to implement, are the sensitivity experiments. So we want to propose different strategies that can be implemented in that zone. So what we want to do is propose the strategies and modify the values of the variables in order to detect the best scenarios that can reduce that risk of cover loss. So if you have any ideas about it, it would be really nice about the strategies that can be implemented. For example, it would be really nice. And that's my contact if you want to say anything about this. Thank you.